42 |
#include "../quant/quant.h" |
#include "../quant/quant.h" |
43 |
#include "../encoder.h" |
#include "../encoder.h" |
44 |
|
|
|
#include "../image/reduced.h" |
|
45 |
#include "../quant/quant_matrix.h" |
#include "../quant/quant_matrix.h" |
46 |
|
|
47 |
MBFIELDTEST_PTR MBFieldTest; |
MBFIELDTEST_PTR MBFieldTest; |
90 |
|
|
91 |
/* Perform DCT */ |
/* Perform DCT */ |
92 |
start_timer(); |
start_timer(); |
93 |
fdct(&data[0 * 64]); |
fdct((short * const)&data[0 * 64]); |
94 |
fdct(&data[1 * 64]); |
fdct((short * const)&data[1 * 64]); |
95 |
fdct(&data[2 * 64]); |
fdct((short * const)&data[2 * 64]); |
96 |
fdct(&data[3 * 64]); |
fdct((short * const)&data[3 * 64]); |
97 |
fdct(&data[4 * 64]); |
fdct((short * const)&data[4 * 64]); |
98 |
fdct(&data[5 * 64]); |
fdct((short * const)&data[5 * 64]); |
99 |
stop_dct_timer(); |
stop_dct_timer(); |
100 |
} |
} |
101 |
|
|
105 |
const uint8_t cbp) |
const uint8_t cbp) |
106 |
{ |
{ |
107 |
start_timer(); |
start_timer(); |
108 |
if(cbp & (1 << (5 - 0))) idct(&data[0 * 64]); |
if(cbp & (1 << (5 - 0))) idct((short * const)&data[0 * 64]); |
109 |
if(cbp & (1 << (5 - 1))) idct(&data[1 * 64]); |
if(cbp & (1 << (5 - 1))) idct((short * const)&data[1 * 64]); |
110 |
if(cbp & (1 << (5 - 2))) idct(&data[2 * 64]); |
if(cbp & (1 << (5 - 2))) idct((short * const)&data[2 * 64]); |
111 |
if(cbp & (1 << (5 - 3))) idct(&data[3 * 64]); |
if(cbp & (1 << (5 - 3))) idct((short * const)&data[3 * 64]); |
112 |
if(cbp & (1 << (5 - 4))) idct(&data[4 * 64]); |
if(cbp & (1 << (5 - 4))) idct((short * const)&data[4 * 64]); |
113 |
if(cbp & (1 << (5 - 5))) idct(&data[5 * 64]); |
if(cbp & (1 << (5 - 5))) idct((short * const)&data[5 * 64]); |
114 |
stop_idct_timer(); |
stop_idct_timer(); |
115 |
} |
} |
116 |
|
|
183 |
const uint16_t * const Zigzag, |
const uint16_t * const Zigzag, |
184 |
const uint16_t * const QuantMatrix, |
const uint16_t * const QuantMatrix, |
185 |
int Non_Zero, |
int Non_Zero, |
186 |
int Sum); |
int Sum, |
187 |
|
int Lambda_Mod); |
188 |
|
|
189 |
/* Quantize all blocks -- Inter mode */ |
/* Quantize all blocks -- Inter mode */ |
190 |
static __inline uint8_t |
static __inline uint8_t |
217 |
|
|
218 |
sum = quant[mpeg](&qcoeff[i*64], &data[i*64], pMB->quant, pParam->mpeg_quant_matrices); |
sum = quant[mpeg](&qcoeff[i*64], &data[i*64], pMB->quant, pParam->mpeg_quant_matrices); |
219 |
|
|
220 |
if(sum && (frame->vop_flags & XVID_VOP_TRELLISQUANT)) { |
if(sum && (pMB->quant > 2) && (frame->vop_flags & XVID_VOP_TRELLISQUANT)) { |
221 |
const uint16_t *matrix; |
const uint16_t *matrix; |
222 |
const static uint16_t h263matrix[] = |
const static uint16_t h263matrix[] = |
223 |
{ |
{ |
236 |
pMB->quant, &scan_tables[0][0], |
pMB->quant, &scan_tables[0][0], |
237 |
matrix, |
matrix, |
238 |
63, |
63, |
239 |
sum); |
sum, |
240 |
|
pMB->lambda[i]); |
241 |
} |
} |
242 |
stop_quant_timer(); |
stop_quant_timer(); |
243 |
|
|
311 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
312 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
313 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * 8; |
|
int32_t cst; |
|
|
int vop_reduced; |
|
314 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
315 |
const IMAGE * const pCurrent = &frame->image; |
const IMAGE * const pCurrent = &frame->image; |
|
transfer_operation_8to16_t * const functions[2] = |
|
|
{ |
|
|
(transfer_operation_8to16_t *)transfer_8to16copy, |
|
|
(transfer_operation_8to16_t *)filter_18x18_to_8x8 |
|
|
}; |
|
|
transfer_operation_8to16_t *transfer_op = NULL; |
|
|
|
|
|
vop_reduced = !!(frame->vop_flags & XVID_VOP_REDUCED); |
|
316 |
|
|
317 |
/* Image pointers */ |
/* Image pointers */ |
318 |
pY_Cur = pCurrent->y + (y_pos << (4+vop_reduced)) * stride + (x_pos << (4+vop_reduced)); |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
319 |
pU_Cur = pCurrent->u + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced)); |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
320 |
pV_Cur = pCurrent->v + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced)); |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
|
|
/* Block size */ |
|
|
cst = 8<<vop_reduced; |
|
|
|
|
|
/* Operation function */ |
|
|
transfer_op = functions[vop_reduced]; |
|
321 |
|
|
322 |
/* Do the transfer */ |
/* Do the transfer */ |
323 |
start_timer(); |
start_timer(); |
324 |
transfer_op(&data[0 * 64], pY_Cur, stride); |
transfer_8to16copy(&data[0 * 64], pY_Cur, stride); |
325 |
transfer_op(&data[1 * 64], pY_Cur + cst, stride); |
transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); |
326 |
transfer_op(&data[2 * 64], pY_Cur + next_block, stride); |
transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); |
327 |
transfer_op(&data[3 * 64], pY_Cur + next_block + cst, stride); |
transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); |
328 |
transfer_op(&data[4 * 64], pU_Cur, stride2); |
transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); |
329 |
transfer_op(&data[5 * 64], pV_Cur, stride2); |
transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); |
330 |
stop_transfer_timer(); |
stop_transfer_timer(); |
331 |
} |
} |
332 |
|
|
344 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
345 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
346 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * 8; |
|
uint32_t cst; |
|
|
int vop_reduced; |
|
347 |
const IMAGE * const pCurrent = &frame->image; |
const IMAGE * const pCurrent = &frame->image; |
348 |
|
|
349 |
/* Array of function pointers, indexed by [vop_reduced<<1+add] */ |
/* Array of function pointers, indexed by [add] */ |
350 |
transfer_operation_16to8_t * const functions[4] = |
transfer_operation_16to8_t * const functions[2] = |
351 |
{ |
{ |
352 |
(transfer_operation_16to8_t*)transfer_16to8copy, |
(transfer_operation_16to8_t*)transfer_16to8copy, |
353 |
(transfer_operation_16to8_t*)transfer_16to8add, |
(transfer_operation_16to8_t*)transfer_16to8add, |
|
(transfer_operation_16to8_t*)copy_upsampled_8x8_16to8, |
|
|
(transfer_operation_16to8_t*)add_upsampled_8x8_16to8 |
|
354 |
}; |
}; |
355 |
|
|
356 |
transfer_operation_16to8_t *transfer_op = NULL; |
transfer_operation_16to8_t *transfer_op = NULL; |
357 |
|
|
|
/* Makes this vars booleans */ |
|
|
vop_reduced = !!(frame->vop_flags & XVID_VOP_REDUCED); |
|
|
|
|
358 |
/* Image pointers */ |
/* Image pointers */ |
359 |
pY_Cur = pCurrent->y + (y_pos << (4+vop_reduced)) * stride + (x_pos << (4+vop_reduced)); |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
360 |
pU_Cur = pCurrent->u + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced)); |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
361 |
pV_Cur = pCurrent->v + (y_pos << (3+vop_reduced)) * stride2 + (x_pos << (3+vop_reduced)); |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
362 |
|
|
363 |
if (pMB->field_dct) { |
if (pMB->field_dct) { |
364 |
next_block = stride; |
next_block = stride; |
365 |
stride *= 2; |
stride *= 2; |
366 |
} |
} |
367 |
|
|
|
/* Block size */ |
|
|
cst = 8<<vop_reduced; |
|
|
|
|
368 |
/* Operation function */ |
/* Operation function */ |
369 |
transfer_op = functions[(vop_reduced<<1) + add]; |
transfer_op = functions[add]; |
370 |
|
|
371 |
/* Do the operation */ |
/* Do the operation */ |
372 |
start_timer(); |
start_timer(); |
373 |
if (cbp&32) transfer_op(pY_Cur, &data[0 * 64], stride); |
if (cbp&32) transfer_op(pY_Cur, &data[0 * 64], stride); |
374 |
if (cbp&16) transfer_op(pY_Cur + cst, &data[1 * 64], stride); |
if (cbp&16) transfer_op(pY_Cur + 8, &data[1 * 64], stride); |
375 |
if (cbp& 8) transfer_op(pY_Cur + next_block, &data[2 * 64], stride); |
if (cbp& 8) transfer_op(pY_Cur + next_block, &data[2 * 64], stride); |
376 |
if (cbp& 4) transfer_op(pY_Cur + next_block + cst, &data[3 * 64], stride); |
if (cbp& 4) transfer_op(pY_Cur + next_block + 8, &data[3 * 64], stride); |
377 |
if (cbp& 2) transfer_op(pU_Cur, &data[4 * 64], stride2); |
if (cbp& 2) transfer_op(pU_Cur, &data[4 * 64], stride2); |
378 |
if (cbp& 1) transfer_op(pV_Cur, &data[5 * 64], stride2); |
if (cbp& 1) transfer_op(pV_Cur, &data[5 * 64], stride2); |
379 |
stop_transfer_timer(); |
stop_transfer_timer(); |
766 |
const uint16_t * const Zigzag, |
const uint16_t * const Zigzag, |
767 |
const uint16_t * const QuantMatrix, |
const uint16_t * const QuantMatrix, |
768 |
int Non_Zero, |
int Non_Zero, |
769 |
int Sum) |
int Sum, |
770 |
|
int Lambda_Mod) |
771 |
{ |
{ |
772 |
|
|
773 |
/* Note: We should search last non-zero coeffs on *real* DCT input coeffs |
/* Note: We should search last non-zero coeffs on *real* DCT input coeffs |
777 |
* helps. */ |
* helps. */ |
778 |
typedef struct { int16_t Run, Level; } NODE; |
typedef struct { int16_t Run, Level; } NODE; |
779 |
|
|
780 |
NODE Nodes[65], Last; |
NODE Nodes[65], Last = { 0, 0}; |
781 |
uint32_t Run_Costs0[64+1]; |
uint32_t Run_Costs0[64+1]; |
782 |
uint32_t * const Run_Costs = Run_Costs0 + 1; |
uint32_t * const Run_Costs = Run_Costs0 + 1; |
783 |
|
|
784 |
/* it's 1/lambda, actually */ |
/* it's 1/lambda, actually */ |
785 |
const int Lambda = Trellis_Lambda_Tabs[Q-1]; |
const int Lambda = (Lambda_Mod*Trellis_Lambda_Tabs[Q-1])>>LAMBDA_EXP; |
786 |
|
|
787 |
int Run_Start = -1; |
int Run_Start = -1; |
788 |
uint32_t Min_Cost = 2<<TL_SHIFT; |
uint32_t Min_Cost = 2<<TL_SHIFT; |
979 |
memset(Out, 0x00, 64*sizeof(*Out)); |
memset(Out, 0x00, 64*sizeof(*Out)); |
980 |
Out[Zigzag[Last_Node]] = Last.Level; |
Out[Zigzag[Last_Node]] = Last.Level; |
981 |
i = Last_Node - Last.Run; |
i = Last_Node - Last.Run; |
982 |
Sum = 0; |
Sum = abs(Last.Level); |
983 |
while(i>=0) { |
while(i>=0) { |
984 |
Out[Zigzag[i]] = Nodes[i].Level; |
Out[Zigzag[i]] = Nodes[i].Level; |
985 |
Sum += abs(Nodes[i].Level); |
Sum += abs(Nodes[i].Level); |